Derived Tree Block Division for Non-Power-of-Two Video Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Intra derived tree-based division modes in video coding result in prediction blocks with side lengths that are not integer powers of 2, leading to inefficient video coding due to unreasonable sub-block divisions, which affect hardware implementation complexity and coding efficiency.
Innovation Solution
Implement a video decoding method that divides prediction blocks with non-integer power side lengths into sub-blocks with integer power side lengths using improved division modes, such as horizontal and vertical derived trees, allowing for inverse quantization and inverse transform to generate reconstructed images efficiently.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If intra derived tree-based division mode is used, then prediction block can be divided into sub-blocks, but the prediction block side length becomes non-integer power of 2 which reduces transform efficiency
Solution Approach 1:
The patent segments the prediction block into multiple sub-blocks with integer power of 2 side lengths. For example, a prediction block of size 8x6 is divided into two sub-blocks: 8x4 and 8x2. This segmentation allows the transform operation to be performed on blocks with standard integer power dimensions, improving transform efficiency while maintaining the flexibility of derived tree-based division modes.
Solution Approach 2:
The patent applies different division strategies to different regions of the prediction block based on their characteristics. By analyzing the residual distribution and frequency content in different regions, the method selects appropriate sub-block divisions that optimize transform efficiency locally, rather than using a uniform division approach for the entire prediction block.
2Adaptability or versatility
If prediction block is divided into sub-blocks with non-integer power side lengths, then division flexibility is improved, but hardware implementation complexity increases
Solution Approach 1:
The patent changes the parameter of sub-block side lengths from non-integer powers of 2 to integer powers of 2. This parameter change simplifies the hardware implementation because integer power dimensions can be handled more efficiently by standard hardware circuits for inverse quantization and inverse transform operations, reducing the complexity of control logic and data processing paths.
3Ease of manufacture
If unreasonable division mode is used for sub-block, then prediction block can be processed, but video coding efficiency is reduced
Solution Approach 1:
The patent incorporates feedback mechanisms to evaluate the coding efficiency of different division modes and adjust the division strategy accordingly. By analyzing the residual distribution and coding performance, the system selects division modes that optimize the balance between processing feasibility and coding efficiency, avoiding unreasonable divisions that would reduce overall performance.
Data Source
AI summary
In a method for video decoding, a bitstream that includes coded information of at least a video image is received. The video image includes a current coding block that adopts a derived tree, the current coding block being divided into a smaller prediction block and a larger prediction block according to the derived tree, the larger prediction block having a side length that is not an integer power of 2. First and second sub coefficient blocks of different sizes for the larger prediction block are decoded from the coded information. The first and second sub coefficient blocks have side lengths that are integer powers of 2, a target division mode for dividing the larger prediction block into the first and second sub coefficient blocks being limited to one or two division modes. A reconstructed image is generated according to the first and second sub coefficient blocks.